We are developing and testing automated techniques for assigning protein structure to novel, uncharacterized sequence, a technique called fold-recognition. Secondary structure predictions are now used with our hidden Markov model (HMM) approach for protein fold recognition, a proceedure called FORESST. The power of this technique has been carefully assessed in a controlled retrospective statistical study. Our study showed that FORESST is more effective than other techniques in finding very distant homologous folds in the range of 12-18% sequence identity. We are currently participating in the fourth community assessment of protein structure prediction techiniques (CASP4) in which this method, augmented with 300 models of protein structural families is being used to predict the structure of novel proteins with as yet unpublished structures. A web-based server for this method has been improved and completely automated to allow for external users to assess the prediction accuracy in a completely objective manner.ABS staff also provided statistical advice and collaboration in areas of statistical analysis, ligand binding data analysis, dose-response curve analysis, repeated-measures ANOVA and MANOVA. Software download for the P-SCAN program was made to the entire Web at version 1.2, and F-SCAN also became available this year. The ABS web server also provides a number of unique sequence analysis services, including secondary structure prediction by the GOR4 algorithm, multiple sequence alignment by CLUSTALW, and various reformatting services.